14 MEMOIRS NATIONAL ACADEMY OF SCIENCES, VOL. XI. 
paratively, made effective use of it in differentiating the agaves of the United States. This 
difference in insertion of the filaments is very marked when a species like Agave barbadensis 
(pl. 38), where they are nearly in the throat, is compared with one like A. sisalana (pl. 114), 
where they become free at about the middle of the tube; but the stamens of Agave are in two 
series, and the two sets are commonly inserted at somewhat different heights in the tube, so 
that in closely related forms the fractional differences that may be observed are not readily 
specified or measured with trustworthy precision. In proportion as the bottom of the tube is 
flattened, also, the insertion of deep-seated filaments may appear to be lowered from the 
middle to the base. The absolute length of the filaments and their ratio to the length of the 
segments of the perianth seem to be fairly characteristic but subject to wide fluctuation; on 
the other hand, the style, which at first is short but, even after the maturity of its stigma, 
continues to grow indefinitely, can be relied on to a very slight extent only as affording specific 
characters. The anthers of related species are usually much alike in general and rather variable 
in particular. 
FRUIT.1 
As in other groups of plants, the fruit and seeds afford characters of a fairly trustworthy 
character. Marked differences separate the large, broad, rather thick-walled capsules of Agave 
Karatto (pl. 15), the elongated, equally thick-walled capsules of A. Dussiana (pl. 29), the smaller, 
nearly round, thinner-walled capsules of A. portoricensis (pl. 78), and the small, round, almost 
papery capsules of A. papyrocarpa (pl. 97). Among the West Indian agaves there may be also 
marked differences in the shape of the base of the capsules (e. g., A. portoricensis, pl. 78, and 
A, papyrocarpa, pl. 97) and of their distal ends (e. g., A. Dussiana, pl. 29, and A. indagatorum, 
pl. 92), even when in size and general proportions they are comparable. Abundant material 
usually shows the constancy of such differences, but starveling fruits (A. anomala, pl. 66) here 
as elsewhere must be taken as uncharacteristic. The seed dimensions of such fruits are not to 
be depended on any more than the measurements of seeds from the extreme top or bottom of 
a normal capsule, though those from the middle of well developed fruits average up pretty 
uniformly ; and the seeds may differ measurably in unrelated species (e. g., A. Brittoniana, 
pl. 98, and A. angustifolia, pl. 109). It must be questioned, until further material is secured, 
whether the small capsules of A. Mhilspaughii (pl. 88) or the narrow capsules of A. cacozela 
(pl. 90) and A. Harrisi (pl. 51), of which few and not obviously normal fruits have been 
observed, are as representative as those of some other species are known to be. 
BULBILS. 
The agaves first known to be bulbiferous, Agave vivipara and A. sobolifera, received their 
specific names because of this fact, and a good deal of confusion has crept into the literature of 
Agave through a misconception as to the constancy of bulbil formation in a given species and 
its prevalence through the genus. The Sisalanae are markedly bulbiferous, and yet A. angus- 
tifolia, which typically shares this trait with its near relatives, sometimes fails to produce bulbils 
when capsules are found abundantly (pl. 107), or defers their production to a period subsequent 
to the ripening of the seed. On the other hand, a few littaeas of the marginate and filiferous 
groups are known to be proliferously viviparous at the end of the spike above the latest flowers; 
Agave attenuata has been shown very recently to be freely bulbiferous sometimes; the lower 
part of the scape of regularly bulbiferous species like A. barbadensis may bear, in the axils 
of the bracts, dense masses of unusually shaped bulbils which develop into as unusual plants 
(pl. C); similar plantlets are often borne on the trunk of polycarpic littaeas or of the 
magueys when they have been prevented from flowering normally; and it may be unwise to 
consider the absence of inflorescence bulbils as of constant diagnostic value even in species 
like A. atrovirens and A. picta, which have not yet been reported to bear them. 
When produced, the bulbils of one species are usually somewhat different from those of 
another, indeed those of widely distinct groups differ greatly; but their differences have been 
little utilized as yet for the characterization of species. However constant it may be, the 
1 Engelmann, Trans. Acad. Sci. St. Louis, vol. 3, 1875, p. 299; Bot. Works, 1887, p- 304. 
